This paper investigates the impact of artificial intelligence on the creative destruction of economic activities in U.S. metropolitan areas. We theorize that the impact of AI is unevenly distributed across regions and that its local impacts can be estimated using variation in exposure to Artificial Intelligence patents. We use a newly constructed dataset linking patents, American Community Survey and O*NET data for 2009-2019 in the U.S.. Preliminary results show that the current development of the U.S. AI sector creates “superstar” AI hubs, meaning that a disproportionate share of economic benefits fall in such MSAs, leaving many more places behind. We argue that the evolving occupational structure in AI-intensive MSAs due to technological change can partially explain the changing economic landscape in the U.S..  MSAs with high AI patenting intensity saw a higher than national average growth in occupations that are complimented by AI compared to less AI-intensive MSAs. However, they have yet seen a stronger decline in occupations that are substituted by AI, probably due to the fact that AI penetration in business still takes time. In addition, the nature of jobs changes faster in MSAs with high AI intensity, moving away from routine cognitive and routine manual tasks towards non-routine tasks that require creative analytical skills or physical dexterity.